Inference on the Edge: The Next Generation of Wearable Silicon for Healthtech
Abstract:
Drawing on direct experience from consumer health wearables at Fitbit and Google, we explore how advances in health sensing are driving a fundamental shift in embedded system architecture. Traditional DSP+MCU approaches are giving way to heterogeneous embedded platforms that combine specialized NPUs and ML models capable of inferring health insights from multimodal physiological signals. We will discuss the increase in complexity for sensing data and how these evolving workloads are changing silicon requirements, from memory architecture, power domains, sensor fusion, and low-power NPU cores.
Biography:
An engineering leader with over 20 years of experience, Pieris Berreitter has spent the last decade leading smartwatch and fitness tracker development at industry leaders Fitbit and Google. At Fitbit Pieris oversaw technology innovation and product development for 8 generations of smartwatches and fitness trackers, and launched health and medical features like basal heart rate, cycle tracking, and ECG Afib detection. At Google, Pieris led the Pixel Watch 2 and Watch 4 programs. Now at HP, Pieris continues to focus on wearables, bringing his passion for technology and on-device compute to the enterprise space.